DPA
2025-05-27
Divergence Point Analysis on Eye-tracking Data
This R script generates timecourse plots of mean fixation proportion, and performs a statistical analysis to determine divergence points between fixations on different objects. It is adapted from the R script in the paper “Analysing data from the psycholinguistic visual‑world paradigm: Comparison of different analysis methods”, Ito & Knoeferle (2023).
The columns in the original eye-tracking data set are:
TOI: trial of interest
Participant: the code for a participant
Timeline: the version of the experiment
Bin: the number of the time bin for one participant and trial
Bin_duration: the duration of each time bin
Competitor: binomially coded fixation data - 1=fixation on competitor object in the current time bin; 0=fixation elsewhere
Distractor: binomially coded fixation data - 1=fixation on distractor object in the current time bin; 0=fixation elsewhere
Target: binomially coded fixation data - 1=fixation on target object in the current time bin; 0=fixation elsewhere
TimeMS: the time bin converted to time in milliseconds
Animacy_C: the animacy of the competitor noun
Animacy_D: the animacy of the distractor noun
Animacy_T: the animacy of the target noun
Determiner: the type of determiner in the trial (Definite or Possessive)
N_Gender_C: the lexical gender of the competitor noun (Masculine or Feminine)
N_Gender_D: the lexical gender of the distractor noun (Masculine or Feminine)
N_Gender_T: the lexical gender of the target noun (Masculine or Feminine)
Phonol_Overlap_C: whether the competitor noun has a minimal pair that only differs in the word-final vowel and lexical gender
Phonol_Overlap_D: whether the distractor noun has a minimal pair that only differs in the word-final vowel and lexical gender
Phonol_Overlap_T: whether the target noun has a minimal pair that only differs in the word-final vowel and lexical gender
ArticleOnset: the onset time (in milliseconds) of the determiner in the audio cue of the current trial
NounOnset: the onset time (in milliseconds) of the noun in the audio cue of the current trial
EndingOnset: the onset time (in milliseconds) of the noun’s final vowel in the audio cue of the current trial
EndingOffset: the offset time (in milliseconds) of the noun’s final vowel in the audio cue of the current trial
Remaining: the remaining time after the offset of the audio cue of the current trial
SpeakerType: the bilingual status of participant - 1 (Adult Spanish Immigrant) or 2 (Heritage Spanish Speaker)
Sex: the sex of the participant (Femenino or Masculino)
Age: the age of the participant
Position: the position of the object in fixation (top left, bottom centre, top right)
Trackloss: whether the fixation is in blank space or off screen, or there is a blink in the current time bin (TRUE or FALSE)
TimeMS_adjusted: TimeMS zeroed at ArticleOnset
ArticleOnset_adjusted: ArticleOnset zeroed at ArticleOnset (thus always 0)
NounOnset_adjusted: NounOnset zeroed at ArticleOnset
EndingOnset_adjusted: EndingOnset zeroed at ArticleOnset
EndingOffset_adjusted: EndingOffset zeroed at ArticleOnset
Remaining_adjusted: Remaining zeroed at ArticleOnset
1. Data preparation
rm(list = ls()) # Clear the workspace
options(scipen = 999) # Turn off scientific notationPackages
require(Rmisc)
require(tidyverse)
require(ggplot2)
require(boot)
require(dplyr)
require(lme4)Package versions
## lme4 Matrix boot lubridate forcats stringr dplyr purrr
## "1.1-37" "1.7-0" "1.3-30" "1.9.4" "1.0.0" "1.5.1" "1.1.4" "1.0.4"
## readr tidyr tibble ggplot2 tidyverse Rmisc plyr lattice
## "2.1.5" "1.3.1" "3.2.1" "3.5.2" "2.0.0" "1.5.1" "1.8.9" "0.22-6"Read the time-course output data of the eyetracker from file
fix.dat = read_delim("/Users/cindyzhang/Desktop/Spanish Gender Project/Spanish Gender Project/data-2025-03-12/df_cleaned.csv", delim=',')
# all the trackloss bins have been removed - in any given row, there is a fixation in one of the 3 AreaFixated Inspect the raw data
summary(fix.dat)## ...1 TOI Participant Timeline
## Min. : 1 Length:732647 Length:732647 Length:732647
## 1st Qu.:183162 Class :character Class :character Class :character
## Median :366324 Mode :character Mode :character Mode :character
## Mean :366324
## 3rd Qu.:549486
## Max. :732647
##
## Bin Bin_duration Competitor Distractor
## Min. : 1.0 Min. : 1.000 Min. :0.0000 Min. :0.0000
## 1st Qu.:181.0 1st Qu.:10.000 1st Qu.:0.0000 1st Qu.:0.0000
## Median :342.0 Median :10.000 Median :0.0000 Median :0.0000
## Mean :338.2 Mean : 9.991 Mean :0.2076 Mean :0.1891
## 3rd Qu.:499.0 3rd Qu.:10.000 3rd Qu.:0.0000 3rd Qu.:0.0000
## Max. :729.0 Max. :10.000 Max. :1.0000 Max. :1.0000
##
## Target TimeMS Animacy_C Animacy_D
## Min. :0.0000 Min. : 10 Length:732647 Length:732647
## 1st Qu.:0.0000 1st Qu.:1810 Class :character Class :character
## Median :1.0000 Median :3420 Mode :character Mode :character
## Mean :0.6033 Mean :3381
## 3rd Qu.:1.0000 3rd Qu.:4990
## Max. :1.0000 Max. :7283
##
## Animacy_T Determiner N_ending_C N_ending_D
## Length:732647 Length:732647 Length:732647 Length:732647
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_ending_T N_Gender_C N_Gender_D N_Gender_T
## Length:732647 Length:732647 Length:732647 Length:732647
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## Phonol_Overlap_C Phonol_Overlap_D Phonol_Overlap_T ArticleOnset
## Length:732647 Mode:logical Length:732647 Min. :2403
## Class :character NA's:732647 Class :character 1st Qu.:2448
## Mode :character Mode :character Median :2479
## Mean :2490
## 3rd Qu.:2524
## Max. :2679
##
## NounOnset EndingOnset EndingOffset Remaining SpeakerType
## Min. :2496 Min. :2591 Min. :2838 Min. :3005 Min. :1.00
## 1st Qu.:2576 1st Qu.:2762 1st Qu.:3020 1st Qu.:3144 1st Qu.:1.00
## Median :2613 Median :2859 Median :3079 Median :3230 Median :2.00
## Mean :2619 Mean :2877 Mean :3107 Mean :3239 Mean :1.54
## 3rd Qu.:2669 3rd Qu.:2972 3rd Qu.:3196 3rd Qu.:3315 3rd Qu.:2.00
## Max. :2807 Max. :3411 Max. :3546 Max. :3656 Max. :2.00
##
## Sex Age Position Trackloss
## Length:732647 Min. :19.00 Length:732647 Mode :logical
## Class :character 1st Qu.:20.00 Class :character FALSE:732647
## Mode :character Median :23.00 Mode :character
## Mean :23.75
## 3rd Qu.:25.00
## Max. :50.00
## NA's :142383
## TimeMS_adjusted ArticleOnset_adjusted NounOnset_adjusted
## Min. :-2670.0 Min. :0 Min. : 70.0
## 1st Qu.: -690.0 1st Qu.:0 1st Qu.:113.0
## Median : 930.0 Median :0 Median :128.0
## Mean : 886.2 Mean :0 Mean :128.3
## 3rd Qu.: 2490.0 3rd Qu.:0 3rd Qu.:144.0
## Max. : 4750.0 Max. :0 Max. :202.0
##
## EndingOnset_adjusted EndingOffset_adjusted Remaining_adjusted
## Min. :170.0 Min. :365.0 Min. : 540.0
## 1st Qu.:276.0 1st Qu.:544.0 1st Qu.: 668.0
## Median :388.0 Median :593.0 Median : 738.0
## Mean :386.3 Mean :616.7 Mean : 748.9
## 3rd Qu.:459.0 3rd Qu.:694.0 3rd Qu.: 818.0
## Max. :773.0 Max. :908.0 Max. :1018.0
##
## GenderMatchTarget_D
## Min. :0.0000
## 1st Qu.:0.0000
## Median :1.0000
## Mean :0.5251
## 3rd Qu.:1.0000
## Max. :1.0000
## head(fix.dat)## # A tibble: 6 × 40
## ...1 TOI Participant Timeline Bin Bin_duration Competitor Distractor
## <dbl> <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl>
## 1 1 Trial01-d… TA105 VersionA 23 10 0 0
## 2 2 Trial01-d… TA105 VersionA 24 10 0 0
## 3 3 Trial01-d… TA105 VersionA 25 10 0 0
## 4 4 Trial01-d… TA105 VersionA 26 10 0 0
## 5 5 Trial01-d… TA105 VersionA 27 10 0 0
## 6 6 Trial01-d… TA105 VersionA 28 10 0 0
## # ℹ 32 more variables: Target <dbl>, TimeMS <dbl>, Animacy_C <chr>,
## # Animacy_D <chr>, Animacy_T <chr>, Determiner <chr>, N_ending_C <chr>,
## # N_ending_D <chr>, N_ending_T <chr>, N_Gender_C <chr>, N_Gender_D <chr>,
## # N_Gender_T <chr>, Phonol_Overlap_C <chr>, Phonol_Overlap_D <lgl>,
## # Phonol_Overlap_T <chr>, ArticleOnset <dbl>, NounOnset <dbl>,
## # EndingOnset <dbl>, EndingOffset <dbl>, Remaining <dbl>, SpeakerType <dbl>,
## # Sex <chr>, Age <dbl>, Position <chr>, Trackloss <lgl>, …Remove rows where the time bins are not 10ms
fix.dat <- fix.dat %>% filter(Bin_duration == 10)Add AreaFixated column
fix.dat <- fix.dat %>%
mutate(AreaFixated = case_when(
Competitor == 1 ~ "Competitor",
Distractor == 1 ~ "Distractor",
Target == 1 ~ "Target"
))Add GenderMatch column
fix.dat <- fix.dat %>%
mutate(GenderMatch = case_when(
N_Gender_D == N_Gender_T ~ "D=T",
N_Gender_D == N_Gender_C ~ "D=C"
))Reshape binomially encoded fixation data to a long format for plotting and analysis
In the original experiment, the eyetracker takes 25 samples in each 50ms bin, and records the proportion of samples that fall in each AOI. Our eyetracker takes 1 sample in each 10ms bin, and records which AOI the sample falls in (we already removed the trackloss bins, so the sample must be in one of the 3 AOIs, and binomially encoded as 0 or 1).
fix.long <- fix.dat %>%
pivot_longer(
cols = c(Target, Competitor, Distractor),
names_to = "AOI",
values_to = "FixP"
)Change Participant, TOI,
AreaFixated, SpeakerType, AOI,
Phonol_Overlap_T and GenderMatch to
factor
fix.long = fix.long %>% mutate_at(vars(Participant, TOI, AreaFixated, SpeakerType, AOI, Phonol_Overlap_T, GenderMatch), as.factor)Merge the non minimal pair values
fix.long$Phonol_Overlap <- fix.long$Phonol_Overlap_T %>% fct_collapse(NonMinPair = c("First Syll-unstressed", "First Syll-stressed"))Inspect the data after factoring
head(fix.long)## # A tibble: 6 × 42
## ...1 TOI Participant Timeline Bin Bin_duration TimeMS Animacy_C Animacy_D
## <dbl> <fct> <fct> <chr> <dbl> <dbl> <dbl> <chr> <chr>
## 1 1 Tria… TA105 VersionA 23 10 230 Animate-… Animate-…
## 2 1 Tria… TA105 VersionA 23 10 230 Animate-… Animate-…
## 3 1 Tria… TA105 VersionA 23 10 230 Animate-… Animate-…
## 4 2 Tria… TA105 VersionA 24 10 240 Animate-… Animate-…
## 5 2 Tria… TA105 VersionA 24 10 240 Animate-… Animate-…
## 6 2 Tria… TA105 VersionA 24 10 240 Animate-… Animate-…
## # ℹ 33 more variables: Animacy_T <chr>, Determiner <chr>, N_ending_C <chr>,
## # N_ending_D <chr>, N_ending_T <chr>, N_Gender_C <chr>, N_Gender_D <chr>,
## # N_Gender_T <chr>, Phonol_Overlap_C <chr>, Phonol_Overlap_D <lgl>,
## # Phonol_Overlap_T <fct>, ArticleOnset <dbl>, NounOnset <dbl>,
## # EndingOnset <dbl>, EndingOffset <dbl>, Remaining <dbl>, SpeakerType <fct>,
## # Sex <chr>, Age <dbl>, Position <chr>, Trackloss <lgl>,
## # TimeMS_adjusted <dbl>, ArticleOnset_adjusted <dbl>, …summary(fix.long)## ...1 TOI Participant
## Min. : 1 Trial30-pos-sapo_0 : 28779 TA131 : 81555
## 1st Qu.:183183 Trial41-def-mujer_0 : 28770 TA124 : 80841
## Median :366326 Trial06-def-brujo_0 : 28491 TA128 : 79215
## Mean :366328 Trial08-def-puerta_0: 28377 TA142 : 76869
## 3rd Qu.:549482 Trial10-pos-oso_0 : 28374 TA130 : 76527
## Max. :732646 Trial29-pos-mosca_0 : 28338 TA137 : 76455
## (Other) :2022921 (Other):1722588
## Timeline Bin Bin_duration TimeMS
## Length:2194050 Min. : 1.0 Min. :10 Min. : 10
## Class :character 1st Qu.:180.0 1st Qu.:10 1st Qu.:1800
## Mode :character Median :342.0 Median :10 Median :3420
## Mean :337.6 Mean :10 Mean :3376
## 3rd Qu.:498.0 3rd Qu.:10 3rd Qu.:4980
## Max. :728.0 Max. :10 Max. :7280
##
## Animacy_C Animacy_D Animacy_T Determiner
## Length:2194050 Length:2194050 Length:2194050 Length:2194050
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_ending_C N_ending_D N_ending_T N_Gender_C
## Length:2194050 Length:2194050 Length:2194050 Length:2194050
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_Gender_D N_Gender_T Phonol_Overlap_C Phonol_Overlap_D
## Length:2194050 Length:2194050 Length:2194050 Mode:logical
## Class :character Class :character Class :character NA's:2194050
## Mode :character Mode :character Mode :character
##
##
##
##
## Phonol_Overlap_T ArticleOnset NounOnset EndingOnset
## First Syll-stressed :636219 Min. :2403 Min. :2496 Min. :2591
## First Syll-unstressed:827898 1st Qu.:2448 1st Qu.:2576 1st Qu.:2762
## Minimal Pair :729933 Median :2479 Median :2613 Median :2859
## Mean :2490 Mean :2619 Mean :2877
## 3rd Qu.:2524 3rd Qu.:2669 3rd Qu.:2972
## Max. :2679 Max. :2807 Max. :3411
##
## EndingOffset Remaining SpeakerType Sex Age
## Min. :2838 Min. :3005 1:1008648 Length:2194050 Min. :19.0
## 1st Qu.:3020 1st Qu.:3144 2:1185402 Class :character 1st Qu.:20.0
## Median :3079 Median :3230 Mode :character Median :23.0
## Mean :3107 Mean :3239 Mean :23.7
## 3rd Qu.:3196 3rd Qu.:3315 3rd Qu.:25.0
## Max. :3546 Max. :3656 Max. :50.0
## NA's :426369
## Position Trackloss TimeMS_adjusted ArticleOnset_adjusted
## Length:2194050 Mode :logical Min. :-2670.0 Min. :0
## Class :character FALSE:2194050 1st Qu.: -690.0 1st Qu.:0
## Mode :character Median : 920.0 Median :0
## Mean : 880.9 Mean :0
## 3rd Qu.: 2480.0 3rd Qu.:0
## Max. : 4750.0 Max. :0
##
## NounOnset_adjusted EndingOnset_adjusted EndingOffset_adjusted
## Min. : 70.0 Min. :170.0 Min. :365.0
## 1st Qu.:113.0 1st Qu.:276.0 1st Qu.:544.0
## Median :128.0 Median :388.0 Median :593.0
## Mean :128.3 Mean :386.3 Mean :616.7
## 3rd Qu.:144.0 3rd Qu.:459.0 3rd Qu.:694.0
## Max. :202.0 Max. :773.0 Max. :908.0
##
## Remaining_adjusted GenderMatchTarget_D AreaFixated GenderMatch
## Min. : 540.0 Min. :0.0000 Competitor: 456042 D=C:1041837
## 1st Qu.: 668.0 1st Qu.:0.0000 Distractor: 415515 D=T:1152213
## Median : 738.0 Median :1.0000 Target :1322493
## Mean : 748.9 Mean :0.5252
## 3rd Qu.: 818.0 3rd Qu.:1.0000
## Max. :1018.0 Max. :1.0000
##
## AOI FixP Phonol_Overlap
## Competitor:731350 Min. :0.0000 NonMinPair :1464117
## Distractor:731350 1st Qu.:0.0000 Minimal Pair: 729933
## Target :731350 Median :0.0000
## Mean :0.3333
## 3rd Qu.:1.0000
## Max. :1.0000
## Compute the mean, SD, SE and CI for FixP for each time
bin, AOI, SpeakerType, Determiner and TOI
fix.means = Rmisc::summarySE(fix.long, measurevar='FixP', groupvars=c('TimeMS_adjusted','SpeakerType', 'AOI', 'Determiner'), na.rm=T)## Warning in qt(conf.interval/2 + 0.5, datac$N - 1): NaNs producedhead(fix.means)## TimeMS_adjusted SpeakerType AOI Determiner N FixP sd se ci
## 1 -2670 1 Competitor Definite 5 0 0 0 0
## 2 -2670 1 Distractor Definite 5 0 0 0 0
## 3 -2670 1 Target Definite 5 1 0 0 0
## 4 -2670 2 Competitor Definite 4 0 0 0 0
## 5 -2670 2 Distractor Definite 4 0 0 0 0
## 6 -2670 2 Target Definite 4 1 0 0 0summary(fix.means)## TimeMS_adjusted SpeakerType AOI Determiner
## Min. :-2670.0 1:4011 Competitor:2793 Length:8379
## 1st Qu.: -900.0 2:4368 Distractor:2793 Class :character
## Median : 850.0 Target :2793 Mode :character
## Mean : 853.9
## 3rd Qu.: 2590.0
## Max. : 4750.0
##
## N FixP sd se
## Min. : 1.0 Min. :0.0000 Min. :0.0000 Min. :0.00000
## 1st Qu.:256.0 1st Qu.:0.1065 1st Qu.:0.2959 1st Qu.:0.01722
## Median :285.0 Median :0.3152 Median :0.4342 Median :0.02610
## Mean :261.9 Mean :0.3333 Mean :0.3775 Mean :0.02597
## 3rd Qu.:303.0 3rd Qu.:0.3711 3rd Qu.:0.4724 3rd Qu.:0.02857
## Max. :364.0 Max. :1.0000 Max. :0.5774 Max. :0.33333
## NA's :132 NA's :132
## ci
## Min. :0.00000
## 1st Qu.:0.03390
## Median :0.05136
## Mean :0.05370
## 3rd Qu.:0.05625
## Max. :1.43422
## NA's :132Compute Mean, etc. that takes Phonol_Overlap into account
fix.means.phonol = Rmisc::summarySE(fix.long, measurevar='FixP', groupvars=c('TimeMS_adjusted','SpeakerType', 'AOI', 'Determiner', 'Phonol_Overlap'), na.rm=T)## Warning in qt(conf.interval/2 + 0.5, datac$N - 1): NaNs producedhead(fix.means.phonol)## TimeMS_adjusted SpeakerType AOI Determiner Phonol_Overlap N FixP sd se
## 1 -2670 1 Competitor Definite Minimal Pair 5 0 0 0
## 2 -2670 1 Distractor Definite Minimal Pair 5 0 0 0
## 3 -2670 1 Target Definite Minimal Pair 5 1 0 0
## 4 -2670 2 Competitor Definite Minimal Pair 4 0 0 0
## 5 -2670 2 Distractor Definite Minimal Pair 4 0 0 0
## 6 -2670 2 Target Definite Minimal Pair 4 1 0 0
## ci
## 1 0
## 2 0
## 3 0
## 4 0
## 5 0
## 6 0summary(fix.means.phonol)## TimeMS_adjusted SpeakerType AOI Determiner
## Min. :-2670.0 1:7941 Competitor:5488 Length:16464
## 1st Qu.: -890.0 2:8523 Distractor:5488 Class :character
## Median : 830.0 Target :5488 Mode :character
## Mean : 832.7
## 3rd Qu.: 2540.0
## Max. : 4750.0
##
## Phonol_Overlap N FixP sd
## NonMinPair :8229 Min. : 1.0 Min. :0.0000 Min. :0.0000
## Minimal Pair:8235 1st Qu.: 92.0 1st Qu.:0.1121 1st Qu.:0.3040
## Median :112.0 Median :0.3084 Median :0.4312
## Mean :133.3 Mean :0.3333 Mean :0.3841
## 3rd Qu.:190.0 3rd Qu.:0.3887 3rd Qu.:0.4739
## Max. :246.0 Max. :1.0000 Max. :0.7071
## NA's :384
## se ci
## Min. :0.0000 Min. :0.0000
## 1st Qu.:0.0254 1st Qu.:0.0502
## Median :0.0348 Median :0.0686
## Mean :0.0374 Mean :0.0783
## 3rd Qu.:0.0461 3rd Qu.:0.0915
## Max. :0.5000 Max. :6.3531
## NA's :384 NA's :384Compute Mean, etc. that takes Phonol_Overlap AND GenderMatch into account
fix.means.fourgroups = Rmisc::summarySE(fix.long, measurevar='FixP', groupvars=c('TimeMS_adjusted','SpeakerType', 'AOI', 'Determiner', 'Phonol_Overlap', 'GenderMatch'), na.rm=T)## Warning in qt(conf.interval/2 + 0.5, datac$N - 1): NaNs producedhead(fix.means.fourgroups)## TimeMS_adjusted SpeakerType AOI Determiner Phonol_Overlap GenderMatch
## 1 -2670 1 Competitor Definite Minimal Pair D=C
## 2 -2670 1 Distractor Definite Minimal Pair D=C
## 3 -2670 1 Target Definite Minimal Pair D=C
## 4 -2670 2 Competitor Definite Minimal Pair D=C
## 5 -2670 2 Distractor Definite Minimal Pair D=C
## 6 -2670 2 Target Definite Minimal Pair D=C
## N FixP sd se ci
## 1 5 0 0 0 0
## 2 5 0 0 0 0
## 3 5 1 0 0 0
## 4 4 0 0 0 0
## 5 4 0 0 0 0
## 6 4 1 0 0 0summary(fix.means.fourgroups)## TimeMS_adjusted SpeakerType AOI Determiner
## Min. :-2670 1:15756 Competitor:10768 Length:32304
## 1st Qu.: -900 2:16548 Distractor:10768 Class :character
## Median : 790 Target :10768 Mode :character
## Mean : 790
## 3rd Qu.: 2470
## Max. : 4750
##
## Phonol_Overlap GenderMatch N FixP
## NonMinPair :16155 D=C:16068 Min. : 1.00 Min. :0.0000
## Minimal Pair:16149 D=T:16236 1st Qu.: 44.00 1st Qu.:0.1143
## Median : 62.00 Median :0.3000
## Mean : 67.92 Mean :0.3333
## 3rd Qu.: 95.00 3rd Qu.:0.4054
## Max. :142.00 Max. :1.0000
##
## sd se ci
## Min. :0.0000 Min. :0.0000 Min. :0.0000
## 1st Qu.:0.3043 1st Qu.:0.0354 1st Qu.:0.0708
## Median :0.4280 Median :0.0489 Median :0.0975
## Mean :0.3828 Mean :0.0516 Mean :0.1126
## 3rd Qu.:0.4764 3rd Qu.:0.0635 3rd Qu.:0.1276
## Max. :0.7071 Max. :0.5000 Max. :6.3531
## NA's :525 NA's :525 NA's :525Compute Mean, etc. that takes GenderMatch into account
fix.means.gendermatch = Rmisc::summarySE(fix.long, measurevar='FixP', groupvars=c('TimeMS_adjusted','SpeakerType', 'AOI', 'Determiner', 'GenderMatch'), na.rm=T)## Warning in qt(conf.interval/2 + 0.5, datac$N - 1): NaNs producedhead(fix.means.gendermatch)## TimeMS_adjusted SpeakerType AOI Determiner GenderMatch N FixP sd se ci
## 1 -2670 1 Competitor Definite D=C 5 0 0 0 0
## 2 -2670 1 Distractor Definite D=C 5 0 0 0 0
## 3 -2670 1 Target Definite D=C 5 1 0 0 0
## 4 -2670 2 Competitor Definite D=C 4 0 0 0 0
## 5 -2670 2 Distractor Definite D=C 4 0 0 0 0
## 6 -2670 2 Target Definite D=C 4 1 0 0 0summary(fix.means.gendermatch)## TimeMS_adjusted SpeakerType AOI Determiner GenderMatch
## Min. :-2670.0 1:7953 Competitor:5501 Length:16503 D=C:8202
## 1st Qu.: -890.0 2:8550 Distractor:5501 Class :character D=T:8301
## Median : 830.0 Target :5501 Mode :character
## Mean : 833.2
## 3rd Qu.: 2550.0
## Max. : 4750.0
##
## N FixP sd se
## Min. : 1.0 Min. :0.0000 Min. :0.0000 Min. :0.0000
## 1st Qu.:127.0 1st Qu.:0.1078 1st Qu.:0.2992 1st Qu.:0.0248
## Median :141.0 Median :0.3101 Median :0.4337 Median :0.0366
## Mean :132.9 Mean :0.3333 Mean :0.3830 Mean :0.0359
## 3rd Qu.:156.0 3rd Qu.:0.3816 3rd Qu.:0.4738 3rd Qu.:0.0406
## Max. :196.0 Max. :1.0000 Max. :0.7071 Max. :0.5000
## NA's :405 NA's :405
## ci
## Min. :0.0000
## 1st Qu.:0.0490
## Median :0.0723
## Mean :0.0754
## 3rd Qu.:0.0803
## Max. :6.3531
## NA's :4052. Plotting
Plot mean fixation proportion to each AOI with standard error, dividing by SpeakerType (ASI and HSS) and Determiner type (Definite and Possessive)
fix.plot = ggplot() +
facet_wrap(~Determiner + SpeakerType, ncol=2,
labeller = labeller(
Determiner = c(
"Definite" = "Definite determiner",
"Possessive" = "Possessive determiner"
),
SpeakerType = c(
"1" = "ASI",
"2" = "HSS"
)
)
) + theme_bw() +
geom_vline(xintercept = 0, linetype = "dashed", color = "black", linewidth = 0.8) +
geom_line(data=fix.means,aes(x=TimeMS_adjusted, y=FixP, group=AOI, colour=AOI, lty=AOI), lwd=1.5) + # TODO use SE or CI here?
geom_ribbon(data=fix.means,aes(x=TimeMS_adjusted,ymin=FixP-se,ymax=FixP+se,color=AOI,fill=AOI), linewidth=.2, alpha=.3, lty="dashed", show.legend=F) +
labs(y="Fixation Proportion", x="Time relative to the target word onset (ms)") +
scale_x_continuous(limits=c(-300, 1400),expand=c(0,0),breaks=seq(-300, 1400, 200)) +
scale_color_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_fill_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_linetype_manual('AOI', values=c("solid","dotted", "dotdash")) +
theme(
text=element_text(size=20),
legend.key.height=unit(.3,"in"),
legend.key.width=unit(.6,"in"),
panel.spacing=unit(1, "in"))
fix.plot## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).
Save the plot
ggsave("plots/fixation_plot.png", width = 20, height = 15, dpi = 300)## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).Plot mean fixation proportion to each AOI with standard error, dividing by SpeakerType (ASI and HSS), Determiner type (Definite and Possessive) and Phonological Overlap (Minimal Pair and Non Minimal Pair)
fix.plot.phonol = ggplot() +
facet_wrap(~Determiner + SpeakerType + Phonol_Overlap, ncol=4,
labeller = labeller(
Determiner = c(
"Definite" = "Definite determiner",
"Possessive" = "Possessive determiner"
),
SpeakerType = c(
"1" = "ASI",
"2" = "HSS"
),
Phonol_Overlap = c(
"Minimal Pair" = "Minimal Pair",
"NonMinPair" = "Non Minimal Pair"
)
)
) + theme_bw() +
geom_vline(xintercept = 0, linetype = "dashed", color = "black", linewidth = 0.8) +
geom_line(data=fix.means.phonol,aes(x=TimeMS_adjusted, y=FixP, group=AOI, colour=AOI, lty=AOI), lwd=1.5) + # TODO use SE or CI here?
geom_ribbon(data=fix.means.phonol,aes(x=TimeMS_adjusted,ymin=FixP-se,ymax=FixP+se,color=AOI,fill=AOI), linewidth=.2, alpha=.3, lty="dashed", show.legend=F) +
labs(y="Fixation Proportion", x="Time relative to the target word onset (ms)") +
scale_x_continuous(limits=c(-300, 1400),expand=c(0,0),breaks=seq(-300, 1400, 200)) +
scale_color_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_fill_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_linetype_manual('AOI', values=c("solid","dotted", "dotdash")) +
theme(
text=element_text(size=20),
legend.key.height=unit(.3,"in"),
legend.key.width=unit(.6,"in"),
panel.spacing=unit(1, "in"))
fix.plot.phonol## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).
Save the plot
ggsave("plots/fixation_plot_phonol.png", width = 40, height = 15, dpi = 300)## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).Plot mean fixation proportion to each AOI with standard error, dividing by SpeakerType (ASI and HSS), Determiner type (Definite and Possessive), Phonological Overlap (Minimal Pair and Non Minimal Pair) and GenderMatch (D=T and D=C)
fix.plot.fourgroups = ggplot() +
facet_wrap(~Determiner + SpeakerType + Phonol_Overlap + GenderMatch, ncol=4,
labeller = labeller(
Determiner = c(
"Definite" = "Definite determiner",
"Possessive" = "Possessive determiner"
),
SpeakerType = c(
"1" = "ASI",
"2" = "HSS"
),
Phonol_Overlap = c(
"Minimal Pair" = "Minimal Pair",
"NonMinPair" = "Non Minimal Pair"
),
GenderMatch = c(
"D=T" = "Distractor = Target",
"D=C" = "Distractor = Competitor"
)
)
) + theme_bw() +
geom_vline(xintercept = 0, linetype = "dashed", color = "black", linewidth = 0.8) +
geom_line(data=fix.means.fourgroups,aes(x=TimeMS_adjusted, y=FixP, group=AOI, colour=AOI, lty=AOI), lwd=1.5) + # TODO use SE or CI here?
geom_ribbon(data=fix.means.fourgroups,aes(x=TimeMS_adjusted,ymin=FixP-se,ymax=FixP+se,color=AOI,fill=AOI), linewidth=.2, alpha=.3, lty="dashed", show.legend=F) +
labs(y="Fixation Proportion", x="Time relative to the target word onset (ms)") +
scale_x_continuous(limits=c(-300, 1400),expand=c(0,0),breaks=seq(-300, 1400, 200)) +
scale_color_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_fill_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_linetype_manual('AOI', values=c("solid","dotted", "dotdash")) +
theme(
text=element_text(size=20),
legend.key.height=unit(.3,"in"),
legend.key.width=unit(.6,"in"),
panel.spacing=unit(1, "in"))
fix.plot.fourgroups## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).
Save the plot
ggsave("plots/fixation_plot_fourgroups.png", width = 40, height = 30, dpi = 300)## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).Plot mean fixation proportion to each AOI with standard error, dividing by SpeakerType (ASI and HSS), Determiner type (Definite and Possessive) and GenderMatch (D=T and D=C)
fix.plot.gendermatch = ggplot() +
facet_wrap(~Determiner + SpeakerType + GenderMatch, ncol=4,
labeller = labeller(
Determiner = c(
"Definite" = "Definite determiner",
"Possessive" = "Possessive determiner"
),
SpeakerType = c(
"1" = "ASI",
"2" = "HSS"
),
GenderMatch = c(
"D=T" = "Distractor = Target",
"D=C" = "Distractor = Competitor"
)
)
) + theme_bw() +
geom_vline(xintercept = 0, linetype = "dashed", color = "black", linewidth = 0.8) +
geom_line(data=fix.means.gendermatch,aes(x=TimeMS_adjusted, y=FixP, group=AOI, colour=AOI, lty=AOI), lwd=1.5) + # TODO use SE or CI here?
geom_ribbon(data=fix.means.gendermatch,aes(x=TimeMS_adjusted,ymin=FixP-se,ymax=FixP+se,color=AOI,fill=AOI), linewidth=.2, alpha=.3, lty="dashed", show.legend=F) +
labs(y="Fixation Proportion", x="Time relative to the target word onset (ms)") +
scale_x_continuous(limits=c(-300, 1400),expand=c(0,0),breaks=seq(-300, 1400, 200)) +
scale_color_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_fill_manual('AOI', values=c("#af2013","#002e65","#ffac00")) +
scale_linetype_manual('AOI', values=c("solid","dotted", "dotdash")) +
theme(
text=element_text(size=20),
legend.key.height=unit(.3,"in"),
legend.key.width=unit(.6,"in"),
panel.spacing=unit(1, "in"))
fix.plot.gendermatch## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).
Save the plot
ggsave("plots/fixation_plot_gendermatch.png", width = 40, height = 15, dpi = 300)## Warning: Removed 1686 rows containing missing values or values outside the scale range
## (`geom_line()`).Combine Competitor and Distractor to create NonTarget columns and reduce the data set to two variables
fix.twovar <- fix.dat %>%
mutate(NonTarget = Competitor + Distractor) %>%
select(-Competitor, -Distractor)
fix.twovar <- fix.twovar %>%
mutate(AreaFixated = ifelse(AreaFixated == "Target", "Target", "NonTarget"))
fix.twovar$AreaFixated <- as.factor(fix.twovar$AreaFixated)Pivot binomially encoded columns
fix.twovar <- fix.twovar %>%
pivot_longer(
cols = c(Target, NonTarget),
names_to = "AOI",
values_to = "FixP"
)
fix.twovar = fix.twovar %>% mutate_at(vars(AOI), as.factor)Merge the non minimal pair values
fix.twovar$Phonol_Overlap <- fix.twovar$Phonol_Overlap_T %>% fct_collapse(NonMinPair = c("First Syll-unstressed", "First Syll-stressed"))
summary(fix.twovar)## ...1 TOI Participant Timeline
## Min. : 1 Length:1462700 Length:1462700 Length:1462700
## 1st Qu.:183183 Class :character Class :character Class :character
## Median :366326 Mode :character Mode :character Mode :character
## Mean :366328
## 3rd Qu.:549482
## Max. :732646
##
## Bin Bin_duration TimeMS Animacy_C
## Min. : 1.0 Min. :10 Min. : 10 Length:1462700
## 1st Qu.:180.0 1st Qu.:10 1st Qu.:1800 Class :character
## Median :342.0 Median :10 Median :3420 Mode :character
## Mean :337.6 Mean :10 Mean :3376
## 3rd Qu.:498.0 3rd Qu.:10 3rd Qu.:4980
## Max. :728.0 Max. :10 Max. :7280
##
## Animacy_D Animacy_T Determiner N_ending_C
## Length:1462700 Length:1462700 Length:1462700 Length:1462700
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_ending_D N_ending_T N_Gender_C N_Gender_D
## Length:1462700 Length:1462700 Length:1462700 Length:1462700
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_Gender_T Phonol_Overlap_C Phonol_Overlap_D Phonol_Overlap_T
## Length:1462700 Length:1462700 Mode:logical Length:1462700
## Class :character Class :character NA's:1462700 Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## ArticleOnset NounOnset EndingOnset EndingOffset Remaining
## Min. :2403 Min. :2496 Min. :2591 Min. :2838 Min. :3005
## 1st Qu.:2448 1st Qu.:2576 1st Qu.:2762 1st Qu.:3020 1st Qu.:3144
## Median :2479 Median :2613 Median :2859 Median :3079 Median :3230
## Mean :2490 Mean :2619 Mean :2877 Mean :3107 Mean :3239
## 3rd Qu.:2524 3rd Qu.:2669 3rd Qu.:2972 3rd Qu.:3196 3rd Qu.:3315
## Max. :2679 Max. :2807 Max. :3411 Max. :3546 Max. :3656
##
## SpeakerType Sex Age Position
## Min. :1.00 Length:1462700 Min. :19.00 Length:1462700
## 1st Qu.:1.00 Class :character 1st Qu.:20.00 Class :character
## Median :2.00 Mode :character Median :23.00 Mode :character
## Mean :1.54 Mean :23.75
## 3rd Qu.:2.00 3rd Qu.:25.00
## Max. :2.00 Max. :50.00
## NA's :284246
## Trackloss TimeMS_adjusted ArticleOnset_adjusted NounOnset_adjusted
## Mode :logical Min. :-2670.0 Min. :0 Min. : 70.0
## FALSE:1462700 1st Qu.: -690.0 1st Qu.:0 1st Qu.:113.0
## Median : 920.0 Median :0 Median :128.0
## Mean : 880.9 Mean :0 Mean :128.3
## 3rd Qu.: 2480.0 3rd Qu.:0 3rd Qu.:144.0
## Max. : 4750.0 Max. :0 Max. :202.0
##
## EndingOnset_adjusted EndingOffset_adjusted Remaining_adjusted
## Min. :170.0 Min. :365.0 Min. : 540.0
## 1st Qu.:276.0 1st Qu.:544.0 1st Qu.: 668.0
## Median :388.0 Median :593.0 Median : 738.0
## Mean :386.3 Mean :616.7 Mean : 748.9
## 3rd Qu.:459.0 3rd Qu.:694.0 3rd Qu.: 818.0
## Max. :773.0 Max. :908.0 Max. :1018.0
##
## GenderMatchTarget_D AreaFixated GenderMatch AOI
## Min. :0.0000 NonTarget:581038 Length:1462700 NonTarget:731350
## 1st Qu.:0.0000 Target :881662 Class :character Target :731350
## Median :1.0000 Mode :character
## Mean :0.5252
## 3rd Qu.:1.0000
## Max. :1.0000
##
## FixP Phonol_Overlap
## Min. :0.0 NonMinPair :976078
## 1st Qu.:0.0 Minimal Pair:486622
## Median :0.5
## Mean :0.5
## 3rd Qu.:1.0
## Max. :1.0
## head(fix.twovar)## # A tibble: 6 × 42
## ...1 TOI Participant Timeline Bin Bin_duration TimeMS Animacy_C Animacy_D
## <dbl> <chr> <chr> <chr> <dbl> <dbl> <dbl> <chr> <chr>
## 1 1 Tria… TA105 VersionA 23 10 230 Animate-… Animate-…
## 2 1 Tria… TA105 VersionA 23 10 230 Animate-… Animate-…
## 3 2 Tria… TA105 VersionA 24 10 240 Animate-… Animate-…
## 4 2 Tria… TA105 VersionA 24 10 240 Animate-… Animate-…
## 5 3 Tria… TA105 VersionA 25 10 250 Animate-… Animate-…
## 6 3 Tria… TA105 VersionA 25 10 250 Animate-… Animate-…
## # ℹ 33 more variables: Animacy_T <chr>, Determiner <chr>, N_ending_C <chr>,
## # N_ending_D <chr>, N_ending_T <chr>, N_Gender_C <chr>, N_Gender_D <chr>,
## # N_Gender_T <chr>, Phonol_Overlap_C <chr>, Phonol_Overlap_D <lgl>,
## # Phonol_Overlap_T <chr>, ArticleOnset <dbl>, NounOnset <dbl>,
## # EndingOnset <dbl>, EndingOffset <dbl>, Remaining <dbl>, SpeakerType <dbl>,
## # Sex <chr>, Age <dbl>, Position <chr>, Trackloss <lgl>,
## # TimeMS_adjusted <dbl>, ArticleOnset_adjusted <dbl>, …Compute mean fixation proportions for the two-variable version
fix.means.twovar = Rmisc::summarySE(fix.twovar, measurevar='FixP', groupvars=c('TimeMS_adjusted','SpeakerType', 'AOI', 'Determiner'), na.rm=T)## Warning in qt(conf.interval/2 + 0.5, datac$N - 1): NaNs producedfix.means.twovar = fix.means.twovar %>% mutate_at(vars(SpeakerType, Determiner), as.factor)
summary(fix.means.twovar)## TimeMS_adjusted SpeakerType AOI Determiner
## Min. :-2670.0 1:2674 NonTarget:2793 Definite :2792
## 1st Qu.: -900.0 2:2912 Target :2793 Possessive:2794
## Median : 850.0
## Mean : 853.9
## 3rd Qu.: 2590.0
## Max. : 4750.0
##
## N FixP sd se
## Min. : 1.0 Min. :0.0000 Min. :0.0000 Min. :0.00000
## 1st Qu.:256.0 1st Qu.:0.2891 1st Qu.:0.3629 1st Qu.:0.02152
## Median :285.0 Median :0.5000 Median :0.4597 Median :0.02706
## Mean :261.9 Mean :0.5000 Mean :0.4136 Mean :0.02904
## 3rd Qu.:303.0 3rd Qu.:0.7109 3rd Qu.:0.4802 3rd Qu.:0.02918
## Max. :364.0 Max. :1.0000 Max. :0.5774 Max. :0.33333
## NA's :88 NA's :88
## ci
## Min. :0.00000
## 1st Qu.:0.04236
## Median :0.05324
## Mean :0.06087
## 3rd Qu.:0.05745
## Max. :1.43422
## NA's :88head(fix.means.twovar)## TimeMS_adjusted SpeakerType AOI Determiner N FixP sd se ci
## 1 -2670 1 NonTarget Definite 5 0 0 0 0
## 2 -2670 1 Target Definite 5 1 0 0 0
## 3 -2670 2 NonTarget Definite 4 0 0 0 0
## 4 -2670 2 Target Definite 4 1 0 0 0
## 5 -2660 1 NonTarget Definite 5 0 0 0 0
## 6 -2660 1 Target Definite 5 1 0 0 0Plot mean fixation proportion vs. time with the two-variable version of the data
fix.plot.twovar = ggplot() +
facet_wrap(~Determiner + SpeakerType, ncol=2,
labeller = labeller(
Determiner = c(
"Definite" = "Definite determiner",
"Possessive" = "Possessive determiner"
),
SpeakerType = c(
"1" = "ASI",
"2" = "HSS"
)
)
) + theme_bw() +
geom_vline(xintercept = 0, linetype = "dashed", color = "black", linewidth = 0.8) +
geom_line(data=fix.means.twovar,aes(x=TimeMS_adjusted, y=FixP, group=AOI, colour=AOI, lty=AOI), lwd=1.5) +
geom_ribbon(data=fix.means.twovar,aes(x=TimeMS_adjusted,ymin=FixP-se,ymax=FixP+se,color=AOI,fill=AOI), linewidth=.2, alpha=.3, lty="dashed", show.legend=F) +
labs(y="Fixation Proportion", x="Time relative to the target word onset (ms)") +
scale_x_continuous(limits=c(-300, 1400),expand=c(0,0),breaks=seq(-300, 1400, 200)) +
scale_color_manual('AOI', values=c("#969798","#ffac00")) +
scale_fill_manual('AOI', values=c("#969798","#ffac00")) +
scale_linetype_manual('AOI', values=c("solid","dotted")) +
theme(
text=element_text(size=20),
legend.key.height=unit(.3,"in"),
legend.key.width=unit(.6,"in"),
panel.spacing=unit(1, "in"))
fix.plot.twovar## Warning: Removed 1124 rows containing missing values or values outside the scale range
## (`geom_line()`).
Save the plot
ggsave("plots/fixation_plot_twovar.png", width = 20, height = 15, dpi = 300)## Warning: Removed 1124 rows containing missing values or values outside the scale range
## (`geom_line()`).3. DPA
Here we will run a by-participant analysis testing the divergence between the Target and NonTarget AOIs.
Select time window and create stratification variables
div.dat = fix.twovar %>%
filter(1400>=TimeMS_adjusted & TimeMS_adjusted>=-300) %>%
dplyr::mutate(StrataVars=paste(Participant, AOI, TimeMS_adjusted, sep='')) %>%
mutate_at(vars(SpeakerType, StrataVars),as.factor) %>%
droplevels()
summary(div.dat)## ...1 TOI Participant Timeline
## Min. : 85 Length:393148 Length:393148 Length:393148
## 1st Qu.:187259 Class :character Class :character Class :character
## Median :372068 Mode :character Mode :character Mode :character
## Mean :369144
## 3rd Qu.:553190
## Max. :732480
##
## Bin Bin_duration TimeMS Animacy_C
## Min. :211.0 Min. :10 Min. :2110 Length:393148
## 1st Qu.:265.0 1st Qu.:10 1st Qu.:2650 Class :character
## Median :307.0 Median :10 Median :3070 Mode :character
## Mean :306.1 Mean :10 Mean :3061
## 3rd Qu.:347.0 3rd Qu.:10 3rd Qu.:3470
## Max. :408.0 Max. :10 Max. :4080
##
## Animacy_D Animacy_T Determiner N_ending_C
## Length:393148 Length:393148 Length:393148 Length:393148
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_ending_D N_ending_T N_Gender_C N_Gender_D
## Length:393148 Length:393148 Length:393148 Length:393148
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
##
## N_Gender_T Phonol_Overlap_C Phonol_Overlap_D Phonol_Overlap_T
## Length:393148 Length:393148 Mode:logical Length:393148
## Class :character Class :character NA's:393148 Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## ArticleOnset NounOnset EndingOnset EndingOffset Remaining
## Min. :2403 Min. :2496 Min. :2591 Min. :2838 Min. :3005
## 1st Qu.:2450 1st Qu.:2576 1st Qu.:2762 1st Qu.:3020 1st Qu.:3144
## Median :2479 Median :2614 Median :2871 Median :3079 Median :3243
## Mean :2491 Mean :2620 Mean :2879 Mean :3108 Mean :3240
## 3rd Qu.:2524 3rd Qu.:2669 3rd Qu.:2974 3rd Qu.:3196 3rd Qu.:3315
## Max. :2679 Max. :2807 Max. :3411 Max. :3546 Max. :3656
##
## SpeakerType Sex Age Position
## 1:183908 Length:393148 Min. :19.00 Length:393148
## 2:209240 Class :character 1st Qu.:20.00 Class :character
## Mode :character Median :23.00 Mode :character
## Mean :23.79
## 3rd Qu.:25.00
## Max. :50.00
## NA's :77128
## Trackloss TimeMS_adjusted ArticleOnset_adjusted NounOnset_adjusted
## Mode :logical Min. :-300.0 Min. :0 Min. : 70.0
## FALSE:393148 1st Qu.: 160.0 1st Qu.:0 1st Qu.:113.0
## Median : 580.0 Median :0 Median :128.0
## Mean : 564.7 Mean :0 Mean :128.4
## 3rd Qu.: 980.0 3rd Qu.:0 3rd Qu.:144.0
## Max. :1400.0 Max. :0 Max. :202.0
##
## EndingOnset_adjusted EndingOffset_adjusted Remaining_adjusted
## Min. :170.0 Min. :365 Min. : 540.0
## 1st Qu.:276.0 1st Qu.:544 1st Qu.: 668.0
## Median :388.0 Median :594 Median : 739.0
## Mean :387.4 Mean :617 Mean : 749.4
## 3rd Qu.:459.0 3rd Qu.:694 3rd Qu.: 818.0
## Max. :773.0 Max. :908 Max. :1018.0
##
## GenderMatchTarget_D AreaFixated GenderMatch AOI
## Min. :0.0000 NonTarget:192712 Length:393148 NonTarget:196574
## 1st Qu.:0.0000 Target :200436 Class :character Target :196574
## Median :1.0000 Mode :character
## Mean :0.5224
## 3rd Qu.:1.0000
## Max. :1.0000
##
## FixP Phonol_Overlap StrataVars
## Min. :0.0 NonMinPair :260806 TA141NonTarget890 : 48
## 1st Qu.:0.0 Minimal Pair:132342 TA141NonTarget960 : 48
## Median :0.5 TA141Target890 : 48
## Mean :0.5 TA141Target960 : 48
## 3rd Qu.:1.0 TA105NonTarget1080: 47
## Max. :1.0 TA105NonTarget1110: 47
## (Other) :392862head(div.dat)## # A tibble: 6 × 43
## ...1 TOI Participant Timeline Bin Bin_duration TimeMS Animacy_C Animacy_D
## <dbl> <chr> <chr> <chr> <dbl> <dbl> <dbl> <chr> <chr>
## 1 85 Tria… TA105 VersionA 324 10 3240 Animate-… Animate-…
## 2 85 Tria… TA105 VersionA 324 10 3240 Animate-… Animate-…
## 3 86 Tria… TA105 VersionA 325 10 3250 Animate-… Animate-…
## 4 86 Tria… TA105 VersionA 325 10 3250 Animate-… Animate-…
## 5 87 Tria… TA105 VersionA 326 10 3260 Animate-… Animate-…
## 6 87 Tria… TA105 VersionA 326 10 3260 Animate-… Animate-…
## # ℹ 34 more variables: Animacy_T <chr>, Determiner <chr>, N_ending_C <chr>,
## # N_ending_D <chr>, N_ending_T <chr>, N_Gender_C <chr>, N_Gender_D <chr>,
## # N_Gender_T <chr>, Phonol_Overlap_C <chr>, Phonol_Overlap_D <lgl>,
## # Phonol_Overlap_T <chr>, ArticleOnset <dbl>, NounOnset <dbl>,
## # EndingOnset <dbl>, EndingOffset <dbl>, Remaining <dbl>, SpeakerType <fct>,
## # Sex <chr>, Age <dbl>, Position <chr>, Trackloss <lgl>,
## # TimeMS_adjusted <dbl>, ArticleOnset_adjusted <dbl>, …Create two subsets based on the type of determiner
div.dat.def <- div.dat %>% filter(Determiner == "Definite")
div.dat.pos <- div.dat %>% filter(Determiner == "Possessive")Running the lmer bootstrap by trial
This ended up yielding mostly NAs due to the small sample size.
Define a function that runs the bootstrap function for all the trials
run_boot_per_trial <- function(toi_id, df, Niter = 100, strata_col = "StrataVars") {
trial_data <- df %>% filter(TOI == toi_id) %>% droplevels()
boot_result <- boot::boot(
data = trial_data,
statistic = boot_L1L2_glm, # your divergence function
strata = trial_data[[strata_col]],
R = Niter
)
# Extract metadata from the first row (assuming Determiner etc. are constant per trial)
trial_info <- trial_data %>%
slice(1) %>%
select(TOI, Determiner, Phonol_Overlap, GenderMatch) # add more if needed
list(boot = boot_result, info = trial_info)
}Run the function that runs the bootstrap function
trial_ids <- unique(div.dat$TOI)
# Store both the bootstrap and metadata
boot_trials <- lapply(trial_ids, function(toi) run_boot_per_trial(toi, div.dat, Niter = 100))
names(boot_trials) <- trial_idsExtract divergence values + metadata into a summary dataframe
# Function to extract means and metadata
extract_div_summary <- function(result) {
bootres <- result$boot
info <- result$info
tibble::tibble(
TOI = info$TOI,
Determiner = info$Determiner,
Phonol_Overlap = info$Phonol_Overlap,
GenderMatch = info$GenderMatchTarget_D,
ASI_divergence = mean(bootres$t[,2], na.rm = TRUE) * 10 - 10,
HSS_divergence = mean(bootres$t[,3], na.rm = TRUE) * 10 - 10
)
}
# Build summary
div_summary_df <- purrr::map_dfr(boot_trials, extract_div_summary)Bootstrap function using glm (generalized linear model)
boot_L1L2_glm = function(original_data, resample_indices){
dat = original_data[resample_indices, ]
# a statistical test at each timepoint for each group
test_g1 = dat %>% # t-test for L1 group
subset(SpeakerType == 1) %>% group_by(TimeMS_adjusted) %>%
dplyr::summarise(t = summary(glm(FixP ~ AOI, family = binomial))$coefficients["AOITarget", "z value"])
test_g2 = dat %>% # t-test for L2 group
subset(SpeakerType == 2) %>% group_by(TimeMS_adjusted) %>%
dplyr::summarise(t = summary(glm(FixP ~ AOI, family = binomial))$coefficients["AOITarget", "z value"])
# return a TRUE/FALSE vector of significant positive t-scores
# (positive means more looks to the target than unrelated)
t_g1 = test_g1$t > 1.96
t_g2 = test_g2$t > 1.96
# create empty vectors to store onsets
onset_g1 = onset_g2 = c()
# find the index of the earliest run of sequential TRUEs; threshold is Nbins
for (i in 1:(length(t_g2)-Nbins)) {
onset_g1[i] = sum(t_g1[i:(i+Nbins-1)]) == Nbins
onset_g2[i] = sum(t_g2[i:(i+Nbins-1)]) == Nbins
}
# find the difference between onsets
delta_g1g2 = which(onset_g2)[1] - which(onset_g1)[1]
# print
# note: the bootstrap returns the indices of the respective timepoints, not absolute times.
# The annotations to the right of each index (e.g. t[,1]) indicate where in the boot object the bootstrapped onset distributions can be found.
c(delta_g1g2, # onset difference L1 vs. L2 t[,1]
which(onset_g1)[1], # onset bin for looks to target L1 t[,2]
which(onset_g2)[1]) # onset bin for looks to target L2 t[,3]
}Run the glm bootstrap on the two subsets by determiner type
bootres_L1L2_glm_def = boot::boot(
data = div.dat.def, # data set to bootstrap
statistic = boot_L1L2_glm, # bootstrap function
strata = div.dat.def$StrataVars, # stratification variable
R = Niter) # number of iterations
bootres_L1L2_glm_pos = boot::boot(
data = div.dat.pos, # data set to bootstrap
statistic = boot_L1L2_glm, # bootstrap function
strata = div.dat.pos$StrataVars, # stratification variable
R = Niter) # number of iterations
bootres_L1L2_glm_def
bootres_L1L2_glm_posSave the glm results
save(bootres_L1L2_glm_def, file="DPA_results.glm.def.rds")
save(bootres_L1L2_glm_pos, file="DPA_results.glm.pos.rds")Extracting and plotting the divergence points
Load the lmer results
load("DPA_results.lmer.def.rds")
load("DPA_results.lmer.pos.rds")Load the glm results
load("DPA_results.glm.def.rds")
load("DPA_results.glm.pos.rds")Define function to extract divergence points and confidence intervals from bootstrapping results
get_divergence_ms <- function(bootres, index) {
# index = 2 for ASI, 3 for HSS
mean_bin <- mean(bootres$t[, index], na.rm = TRUE)
ci_bin <- boot::boot.ci(bootres, index = index, type = "perc")$percent[4:5]
mean_ms <- (mean_bin - 1) * 10
ci_ms <- (ci_bin - 1) * 10
list(mean = mean_ms, ci = ci_ms)
}Run the function on all the results
# Definite determiner
def_asi <- get_divergence_ms(bootres_L1L2_lmer_def, index = 2)
def_hss <- get_divergence_ms(bootres_L1L2_lmer_def, index = 3)
# Possessive determiner
pos_asi <- get_divergence_ms(bootres_L1L2_lmer_pos, index = 2)
pos_hss <- get_divergence_ms(bootres_L1L2_lmer_pos, index = 3)Create dataframe for plotting
div_df <- tibble::tibble(
Determiner = c("Definite", "Definite", "Possessive", "Possessive"),
SpeakerType = c("1", "2", "1", "2"),
TimeMS = c(def_asi$mean, def_hss$mean, pos_asi$mean, pos_hss$mean),
CI_low = c(def_asi$ci[1], def_hss$ci[1], pos_asi$ci[1], pos_hss$ci[1]),
CI_high = c(def_asi$ci[2], def_hss$ci[2], pos_asi$ci[2], pos_hss$ci[2])
)
div_df = div_df %>% mutate_at(vars(SpeakerType, Determiner), as.factor)
div_df <- div_df %>%
mutate(label = paste0(TimeMS, " ms"))Add divergence points and their confidence intervals to the two-variable plots
fix.plot.twovar +
geom_point(data = div_df,
aes(x = TimeMS, y = 0.48),
shape = 18, size = 4, color = "black") +
geom_errorbarh(data = div_df,
aes(xmin = CI_low, xmax = CI_high, y = 0.48),
height = 0.02, color = "black", linewidth = 0.6) +
geom_text(data = div_df,
aes(x = TimeMS - 50, y = 0.55, label = label),
size = 10, fontface="bold")## Warning: Removed 1124 rows containing missing values or values outside the scale range
## (`geom_line()`).
Save the plots
ggsave("plots/fixation_plot_twovar_divpts.png", width = 20, height = 15, dpi = 300)## Warning: Removed 1124 rows containing missing values or values outside the scale range
## (`geom_line()`).4. Group comparison
We can compute a p-value for the group comparison by creating a bootstrap distribution of the null hypothesis. To do this, we will first pool the original data from both groups, randomly assign group labels, and then estimate a difference in divergence point. We will repeat this many times to obtain a distribution of divergence point that could be expected under the null hypothesis. We will obtain the p-value by calculating the proportion of samples from this null distribution that are larger than the observed difference in divergence point in the empirical data.
Below, we will calculate a p-value for the group comparison (ASI vs. HSS group). We will first define a bootstrap function.
Define the bootstrap function for group comparison (only lmer so far)
boot_L1L2_pval = function(original_data, resample_indices){
dat_resample = original_data[resample_indices, ] # resample the data
dat = dat_resample %>%
group_by(Participant, AOI, TimeMS_adjusted) %>%
transform(SpeakerType=sample(SpeakerType,replace=F)) %>% # randomly assign group labels
ungroup() #%>%
# a statistical test at each timepoint for each group
test_g1 = dat %>% # t-test for L1 group
subset(SpeakerType == 1) %>% group_by(TimeMS_adjusted) %>%
dplyr::summarise(t = summary(lmer(FixP ~ AOI + (1|Participant)))$coefficients["AOITarget", "t value"])
test_g2 = dat %>% # t-test for L2 group
subset(SpeakerType == 2) %>% group_by(TimeMS_adjusted) %>%
dplyr::summarise(t = summary(lmer(FixP ~ AOI + (1|Participant)))$coefficients["AOITarget", "t value"])
# return a TRUE/FALSE vector of significant positive t-scores
# (positive means more looks to the target than unrelated)
t_g1 = test_g1$t > 1.96
t_g2 = test_g2$t > 1.96
# create empty vectors to store onsets
onset_g1 = onset_g2 = c()
# find the index of the earliest run of 4 sequential TRUEs
for (i in 1:(length(t_g2)-Nbins)) {
onset_g1[i] = sum(t_g1[i:(i+Nbins-1)]) == Nbins
onset_g2[i] = sum(t_g2[i:(i+Nbins-1)]) == Nbins
}
# find the difference between onsets
delta_g1g2 = which(onset_g2)[1] - which(onset_g1)[1]
# print
delta_g1g2 # onset difference L1 vs. L2 t[,1]
}Run the lmer bootstrap on the two subsets by determiner type
bootres_L1L2_pval_def = boot::boot(
data = div.dat.def, # data set to bootstrap
statistic = boot_L1L2_pval, # bootstrap function
strata = div.dat.def$StrataVars, # stratification variable
R = Niter) # number of iterations
bootres_L1L2_pval_pos = boot::boot(
data = div.dat.pos, # data set to bootstrap
statistic = boot_L1L2_pval, # bootstrap function
strata = div.dat.pos$StrataVars, # stratification variable
R = Niter) # number of iterations
bootres_L1L2_pval_def
bootres_L1L2_pval_posSave the group comparison results
save(bootres_L1L2_pval, file="DPA_results.pval.def.rds")
save(bootres_L1L2_pval, file="DPA_results.pval.pos.rds")If we already have saved results, we can load them.
load("DPA_results.pval.def.rds")
load("DPA_results.pval.pos.rds")Compute the p-values
pval.def <- round(mean(bootres_L1L2_pval_def$t[,1]>=bootres_L1L2_lmer_def$t0[1], na.rm=T), 5)
pval.pos <- round(mean(bootres_L1L2_pval_pos$t[,1]>=bootres_L1L2_lmer_pos$t0[1], na.rm=T), 5)
pval.def
pval.pos